Pluto and Eris are icy dwarf planets with nearly identical sizes, comparable densities and similar surface compositions as revealed by spectroscopic studies 1,2 . Pluto possesses an atmosphere whereas Eris does not; the difference probably arises from their differing distances from the Sun, and explains their different albedos 3 . Makemake is another icy dwarf planet with a spectrum similar to Eris and Pluto 4 , and is currently at a distance to the Sun intermediate between the two. Although Makemake's size (1,420 6 60 km) and albedo are roughly known 5,6 , there has been no constraint on its density and there were expectations that it could have a Plutolike atmosphere 4,7,8 . Here we report the results from a stellar occultation by Makemake on 2011 April 23. Our preferred solution that fits the occultation chords corresponds to a body with projected axes of 1,430 6 9 km (1s) and 1,502 6 45 km, implying a V-band geometric albedo p V 5 0.77 6 0.03. This albedo is larger than that of Pluto, but smaller than that of Eris. The disappearances and reappearances of the star were abrupt, showing that Makemake has no global Pluto-like atmosphere at an upper limit of 4-12 nanobar (1s) for the surface pressure, although a localized atmosphere is possible. A density of 1.7 6 0.3 g cm 23 is inferred from the data. Stellar occultations allow detection of very tenuous atmospheres and can provide accurate sizes and albedos 9,10,11,3,12 , so we embarked on a programme of predicting and observing occultations by (136472) Makemake, also known as 2005 FY 9 . The occultation of the faint star NOMAD 1181-0235723 (with magnitude m R 5 18.22, where NOMAD is the Naval Observatory Merged Astronomic Dataset) was predicted in 2010 by methods similar to those used to predict occultations by several large bodies 13 , but refined as shown in Supplementary Information section 1. We arranged a campaign involving 16 telescopes, listed in Supplementary Table 1. The occultation was successfully recorded from seven telescopes, listed in Table 1, at five sites. From the images obtained, we made photometric measurements as a function of time (light curves).The light curves of the occultation are shown in Fig. 1. Fitting synthetic square-well models to the light curves yielded the disappearance and reappearance times of the star (Table 1), from which we calculate one chord in the plane of the sky for each site (see Supplementary Information section 3). On the basis of analyses of the light curves, taking into account the cycle time between the images and the dispersion of the data, we deduce that there were no secondary occultations, so we can reject the existence of a satellite larger than about 200 km in diameter in the areas sampled by the chords. The result is consistent with a deep-image survey that did not find any satellites 16 . The chords can be fitted with two shape models (Fig. 2). Our preferred shape, which is compatible with our own and other observations (see Supplementary Information section 8), corresponds to an elliptical object ...